In a memory, an access performance or cost per capacity is different depending on a type or a model even having the same type. Typically, as the access performance becomes higher, the cost per capacity becomes higher. Therefore, the type or the model of a memory is selected in consideration of a balance between the desired access performance and the desired capacity, and the upper limit of the cost.
From such a viewpoint, it is conceived that a plurality of memories having different access performances is utilized effectively. For example, a technology is known in which memories in a storage device are layered in accordance with the access performances, and pieces of data are rearranged between the layers. In such a technology, data having a high access frequency is arranged in a memory having a high performance, and data having a low access frequency is arranged in a memory having a low performance.
In addition, the following technology has been proposed for layering of memories. For example, a storage system has been proposed in which data of a second storage layer is mirrored to a first storage layer, and a read command is caused to be transmitted to the second storage layer when a factor measurement value of quality of service (QoS) is smaller than a specific threshold value.
In addition, the following computing system has also been proposed. In such a computing system, an identical file is created in two memories having different classes, and data is read from the memory in a high class in response to a read request. In addition, when a file is broken at the time of access, the other file is copied to another memory of the same class. Japanese Laid-open Patent Publication No. 2011-165164 and Japanese Laid-open Patent Publication No. 04-107750 are the related arts.